Robust Deepfake Detection: Mitigating Spatial Attention Drift via Calibrated Complementary Ensembles
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915964631646208 |
|---|---|
| author | Le-Phan, Minh-Khoa Le, Minh-Hoang Do, Trong-Le Tran, Minh-Triet |
| author_facet | Le-Phan, Minh-Khoa Le, Minh-Hoang Do, Trong-Le Tran, Minh-Triet |
| contents | Current deepfake detection models achieve state-of-the-art performance on pristine academic datasets but suffer severe spatial attention drift under real-world compound degradations, such as blurring and severe lossy compression. To address this vulnerability, we propose a foundation-driven forensic framework that integrates an extreme compound degradation engine with a structurally constrained, multi-stream architecture. During training, our degradation pipeline systematically destroys high-frequency artifacts, optimizing the DINOv2-Giant backbone to extract invariant geometric and semantic priors. We then process images through three specialized pathways: a Global Texture stream, a Localized Facial stream, and a Hybrid Semantic Fusion stream incorporating CLIP. Through analyzing spatial attribution via Score-CAM and feature stability using Cosine Similarity, we quantitatively demonstrate that these streams extract non-redundant, complementary feature representations and stabilize attention entropy. By aggregating these predictions via a calibrated, discretized voting mechanism, our ensemble successfully suppresses background attention drift while acting as a robust geometric anchor. Our approach yields highly stable zero-shot generalization, achieving Fourth Place in the NTIRE 2026 Robust Deepfake Detection Challenge at CVPR. Code is available at https://github.com/khoalephanminh/ntire26-deepfake-challenge. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_25889 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Robust Deepfake Detection: Mitigating Spatial Attention Drift via Calibrated Complementary Ensembles Le-Phan, Minh-Khoa Le, Minh-Hoang Do, Trong-Le Tran, Minh-Triet Computer Vision and Pattern Recognition Current deepfake detection models achieve state-of-the-art performance on pristine academic datasets but suffer severe spatial attention drift under real-world compound degradations, such as blurring and severe lossy compression. To address this vulnerability, we propose a foundation-driven forensic framework that integrates an extreme compound degradation engine with a structurally constrained, multi-stream architecture. During training, our degradation pipeline systematically destroys high-frequency artifacts, optimizing the DINOv2-Giant backbone to extract invariant geometric and semantic priors. We then process images through three specialized pathways: a Global Texture stream, a Localized Facial stream, and a Hybrid Semantic Fusion stream incorporating CLIP. Through analyzing spatial attribution via Score-CAM and feature stability using Cosine Similarity, we quantitatively demonstrate that these streams extract non-redundant, complementary feature representations and stabilize attention entropy. By aggregating these predictions via a calibrated, discretized voting mechanism, our ensemble successfully suppresses background attention drift while acting as a robust geometric anchor. Our approach yields highly stable zero-shot generalization, achieving Fourth Place in the NTIRE 2026 Robust Deepfake Detection Challenge at CVPR. Code is available at https://github.com/khoalephanminh/ntire26-deepfake-challenge. |
| title | Robust Deepfake Detection: Mitigating Spatial Attention Drift via Calibrated Complementary Ensembles |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.25889 |